fwiw lucene 3.6 is in trunk and it supports jdk7 (so we can switch
opengrok to jdk7 slowly too)
so it might be a good test to check a trunk build if it is a lucene leak
or opengrok bug
obviously tracing this and checking syscalls would be the best, but if
only you can reproduce, then it can be your call only
hmm?
L
On 23.5.2012 10:12, Knut Anders Hatlen wrote:
"Gandhi, Shailey"<shailey.gan...@morganstanley.com> writes:
Please help me with this error:
Error:
Directory ‘/data/opengrok/Indexes/index/3dframework_11.0.0’ exists and
is a directory, but cannot be listed:list() returned null.
This comes up every now and then and I am unable to figure out why.
I'm not sure why you're seeing this. Since it works sometimes, I assume
it's not a problem with file permissions on the directory (but still
worth checking out just in case).
The specification for java.io.File.list() says that it returns null if
"this abstract pathname does not denote a directory, or if an I/O error
occurs." So it could be, well, almost anything...
Might be worth checking if there's a file descriptor leak too (using
pfiles, lsof, or similar tools), which could prevent the directory from
being read. That would explain why it's only seen sometimes.
When we eventually move to Java 7, we should replace calls to
java.io.File.list() with calls to
java.nio.file.Files.newDirectoryStream(), which will throw an
IOException with details about what went wrong.
But I think the specific call that generates the message you saw, comes
from Lucene, so it's not under OpenGrok's control.
_______________________________________________
opengrok-discuss mailing list
opengrok-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opengrok-discuss
